Achieving 24p (film look) from HDV source?
 

I'm doing some experiments with rendering HDV footage to a "film look" 24p. I've had some luck, but am not quite getting the quality I want.

What I've done is this:

1) Capture M2T using HDVSplit
2) Bring into Vegas using the DV Widescreen 24p template
3) Deinterlace method set to "interpolate"
4) Apply an "s" shaped color curve
5) Turn *off* smart resample (otherwise there were lots of ghost images in moving areas)
6) Render using MPEG2 WS 24p template

This looks OK, but rather blocky and not sharp, as if something is not good with the scaling.

I've had better results using Bob deinterlacing in Virtualdub (to get 60p at half resolution), but that is a very tedious manual process.

Is there anything else I can be doing? Setting video render quality to "Best" seemed to help, but not massively.

Stop working in an SD timeline. Your results will be much better.

"Turn *off* smart resample (otherwise there were lots of ghost images in moving areas)"

Does this occur with smart resample? First I've heard of this....

Yep, here are a couple of shots of the same frame. Settings were identical in both, except for the resample: HDV project template, deinterlace method set to interpolate, MPEG2 WS 24p render template, etc.

This one is with resample *disabled*: http://www.villamil.org/images/film-interpolate.jpg

This one is with resample *enabled*: http://www.villamil.org/images/film-resample.jpg

Note how the frame with resampling enabled seems to be an average of three frames, so there is triple image of the bird's wings. The one without smart resampling is very clean.

For completeness, here is the same frame using deinterlacing set to blend, with no resampling. Note the image is doubled, not tripled as with resampling.

http://www.villamil.org/images/film-blend.jpg

Thanks for posting the comparsion shots....interesting as I have never experienced this....hopefully others will chime in....

I don't experience this either. Are these shots post-render? or during editing?

Post-render. Those are frame grabs from the rendered MPEG2.

The only change between the film-interpolate and film-resample was that smart resample was disabled for film-interpolate. Same render template was used for all files.

Istanbul? Looking across the Bosporus to Sultanahmet and the Blue Mosque / Aya Sophia?

Feel free to send me an email via the forum if you're interested in sharing experiences (footage?) on shooting in Turkey.

Sorry, no clue on the sampling/render issue.

Yes indeed! I have about 15 hours of footage from that trip, including the total eclipse...

Will e-mail you.

Yes, this made a huge difference. Does Vegas rescale footage to fit the project properties, leading to two scaling operations (HDV to project size, then project size to render size) and hence loss of quality?

I also tried applying minimal blur as well before render. It looks OK and leads to less blockiness, but the minimal blur setting in Vegas seems a bit too much.

Maybe can use the convolution filter to define an alternative tiny blur?

Thanks for comments, everyone.

I'm currently getting best results from this workflow:

1. Capture with HDVSplit
2. Bring into Vegas using HDV 60i project template
3. Set deinterlace method to "interpolate"
4. Disable resampling
5. Add "s" shaped color curve (lots of room to experiment here...)
6. Add custom blur made with convolution filter, for less blurring than Gaussian blur at minimum setting
7. Render using MPEG2 24p template, except change video render quality to "best"

Any other thoughts/ideas?

Hi, sorry I don't have anything to offer you (you seem to be the one helping me), but I was wondering what the purpose of the blur step was?

Without the blur step, things end up looking a bit "pixelated". I think a bit of blurring reduces the amount of detail in the image, and makes life easier for the MPEG2 encoder.

The motion problems I see with DV with the wings of birds, is not with their general motion of flight or their bodies. Those come out pretty clear. But, it's their wings, that are moving much faster, that register in multiple positions, as you watch real-time playback. This is from 1st-generation tapes, that haven't been run through any post-processing.

At what framerate were these acquired? 50i? 25p? 24p?
Stills don't tell much of the story, but of course you're going to see some frame redundancy on something like this. I'm starting to lose the thread, which might be pointing to lack of sleep, but are you starting at a higher framerate resampling TO 24p? or starting with 24p trying to get rid of the redundant/echoed frames?